A complex process of change is underway in the Central Texas English GOOSE vowel. Koops (2010) describes it as the simultaneous operation of two qualitatively different fronting processes. One is a feature of traditional Southern American English (SAE) while the other results from a recent innovation in mainstream American English (MAE). In this paper, we examine a corpus of digital and digitized recordings of Central Texas English (TxE) speech, spanning 30 years in real and 100 in apparent time, to determine the distribution of these two variants of GOOSE throughout the social spectrum. We test two methods for modeling sub-phonemic variation in our dependent variable and perform regression analysis of both linguistic and social factors conditioning this variation. We find, in particular, that women favor the more prestigious MAE variant, suggesting a sound change in progress that aligns TxE more closely with MAE.